New 5D Raw developments - plus my low light comparison with Blackmagic Cinema Camera

[leo]

Yeah, just a few days ago Andrew was saying it would be highly unlikely that Canon's upcoming DSLR cameras would be capable of a 10 second raw burst, and now on the "old cameras" we have a 40 second limit when shooting continuous raw, and that's because of the file system, not the camera! :)

 

I too was under the impression they can't do it, processing power/buffer.. something, it seems we know nothing.

This can't be of much use for professional use ( ML new hack ) but it sure gets the market excited for the next best thing.

But I am pissed tehnology changes this fast, I get they wanna live too ( Canon,Sony,Nikon,Panasonic.. ) but for god sake  every year something major happens, better ISO, better dynamic, crazy black magic stuff, you just can't keep up with the spending money on gear if you want to offer the best.

I moved from camcorders to DSLR's this year and messed up my whole profit for this year, and the spending just started, I need lenses, good audio, now it seems in the near future I might need at least 2k, what the hell.

I guess I can just ignore it and do with what I have but it doesn't work like that, some new kid on the block enters the market and goes for the latest stuff, maybe he is bad but once he advertise 2k/4k I get the same requests from my clients too.

I tell you this, 2k and 4k to the masses is coming faster than expected, some of us are slaves to constant upgrades rather than make a decent living from event videography.

[Bruno]

This can't be of much use for professional use ( ML new hack ) but it sure gets the market excited for the next best thing.

 

It's been what, a week or two? It's still an alpha, I wouldn't rule out a version that is totally useable under any shooting conditions in the (very) near future.

[Yondaime]

Hi, when the 5D mark III comes out, many spokes about its ability in video mode to bin the pixel togheter 3 at time.. Is this the reason of the execellent performance in low-light and the reason for the near-absence

 of moire and aliasing..? This means, if i'm right, that neither the 6D nor the older 5D MkII will reach in Raw the same quality of the MkIII in FF mode..

 

[Ilias Giarimis]

Just wondering, was the lighting exactly the same for both images?

 

If you look at that strong highlight on the round base of the desklamp, you'll see that the BMCC can completely recover it, while the 5D3 cannot.  This with the "Whites" and "Highlights" slider pulled all the way down to recover as much as possible. 

 

I'd rather shoot 5D3 RAW because of the sensor size, but if the lighting was the same for both images, it really shows the tremendous dynamic range that the BMCC is capable of.

 

This happens because BMCC's RAW is just at base ISO with no analog or digital amplification and with just an exif tag (look for "baseline exposure" in exif) giving the info to the a raw converter which makes use of this tag (ACR/LR) that a 3.4 stop "exposure" increase is needed to render the shot at correct lightness.

 

If we open the BMCC raw with a converter like Rawtherapee which does not make use of this tag the photo shows as 3.4 stop underexposed.

 

Meanwhile Canon's DNG is about 0.5 stops over exposed, with average raw green levels at 2.4 stops lower than saturation instead of the usual 3.0 .. this gave an advantage regarding noise at the dark areas at the expense of burned highlights.

 

So the BMCC has about 4 stops greater headroom for highlights

[Sean Cunningham]

Not totally insignificant too is that the BMCC is also writing 12bit log from a 16bit linear capture.  

[Ilias Giarimis]

Not totally insignificant too is that the BMCC is also writing 12bit log from a 16bit linear capture.  

 

Indeed, this 12bit log raw is a nice implementation gives better elasticity than 14bit linear and it became a necessity in the absense of analog amplification. To withstand 5-6 stops digital amplification a file must have excessive bit depth .. it's the same situation as with Medium Format raws with no analog ISO but 16bit linear data. 

[Andrew Reid]

Hi, when the 5D mark III comes out, many spokes about its ability in video mode to bin the pixel togheter 3 at time.. Is this the reason of the execellent performance in low-light and the reason for the near-absence

 of moire and aliasing..? This means, if i'm right, that neither the 6D nor the older 5D MkII will reach in Raw the same quality of the MkIII in FF mode..

 

This is what I believe.

 

Is it correct? Time will tell.

[peederj]

Whatever gain setting you have at the sensor can be compensated for automatically to adjust it to your bit depth. They know the level of the sensor noise floor and they can just keep that near the LSB, meaning the sensels will saturate before any additional headroom in the bit depth can be taken advantage of. An implementation that doesn't do this is just wasting a lot of bandwidth, or wasting the actual DR of its sensor...the sensor DR is the limiting factor and getting the specific gain mapping right once and for all isn't hard.

[Yondaime]

This is what I believe.

 

Is it correct? Time will tell.

One way to check this assumption by video test with only one model running this new experimental ML firmware, could be to compare the 1920*1080 FF mode vs 1:1 1920*1080 crop-mode in low-light...?  ^_^

[amanieux]

It isn't that the focus is off. The background is more in focus on the BMCC because of the smaller sensor and less shallow DOF. That ain't hard to understand surely.

 

I used 24mm F1.4 on the BMCC, and 50mm F1.4 on the 5D Mark III to match field of view more closely.

excuse me to ask a candid question but can the DOF in the background be identical on both camera if you also used the 50mm on the BMCC but moved the BMCC further back to match the field of view of the 5dmk3 with the 50mm ? 

[chauffeurdevan]

Not totally insignificant too is that the BMCC is also writing 12bit log from a 16bit linear capture.  

 

The BMCC sCMOS have a dynamic range quoted at >16,000:1, 14bit is 16,384. So it is not 16 bit unless there is 2bit of noise.

 

http://www.scmos.com/files/high/scmos_white_paper_8mb.pdf

[Sean Cunningham]

Seriously?  It's already established the original BMCC sensor captures 16bit linear, this is reduced to 12bit log (unpacking back to 16bit once loaded into Resolve).

 

 

edit: sorry, my bad, the sensor is capturing more than 16bits.  It takes that, reduces to 16bit lin, then from that writes a 12bit log file to disc.

[chauffeurdevan]

Seriously?  It's already established the original BMCC sensor captures 16bit linear, this is reduced to 12bit log (unpacking back to 16bit once loaded into Resolve).

 

 

edit: sorry, my bad, the sensor is capturing more than 16bits.  It takes that, reduces to 16bit lin, then from that writes a 12bit log file to disc.

 

 

I read more in detail the sCMOS whitepaper, and I was wrong, so are you.

 

The sensor sampling is done by two ADC - one with high gain amplifier, one with low gain amplifier. Each ADC output 11bit. Two 11bits doesn't give neither 22bits (unless it is MSB and LSB - but it is impossible in this case, amplifier/ADC are linear) The only linear signal that could be created (over 11bit) is a 12bit signal - greatest value of a 11 bit unsigned integer is 2048, 2048+2048=4096=12bit. Over that, it is not linear anymore... but logarithmic. If the high gain was +18db, you need to multiply its ADC output by 3. I will not go deeper, but hope you understand...

 

However, from the measured specs, a ratio of 16,000:1 is best represented by a 14bit linear value. +/-13 stops is best represented by a 14-bit linear value. However I think that BMCC did a better choice of choosing 12bit log instead of 14bit linear.